Enable AVX2 libs, speed flags
Before: 42.94 Mpix/sec | Now: 62.34 Mpix/sec | Now (non-avx): 52.18 Mpix/sec
- The postprocessing_benchmark binary from -utils with assests from openbenchmarking.org was used for benchmarking
- PGO not tested due to no testsuite
- LLVM toolchain not tested as libraw wasn't happy with using llvm's openmp (as opposed to gcc's)
- LTO made performance slightly worse and binaries slightly bigger
- Prefer 256 vector width didn't make a difference over 128.